Published: September 22, 2011 r2011 American Chemical Society 4282 dx.doi.org/10.1021/nl202279z | Nano Lett. 2011, 11, 4282–4287 LETTER pubs.acs.org/NanoLett Contactless Measurement of Surface Dominated Recombination in Gold- and Aluminum-Catalyzed Silicon Vapor Liquid Solid Wires Brian A. Bryce,* ,† Mark C. Reuter, § Brent A. Wacaser, § and Sandip Tiwari ‡ † School of Applied and Engineering Physics and ‡ School of Electrical and Computer Engineering, Cornell University, Ithaca, New York 14853, United States § IBM Research Division, T. J. Watson Research Center, Yorktown Heights, New York 10598, United States b S Supporting Information S emiconductor devices rely on material purity for their properties. The basis for nearly all semiconductor devices is the introduction of a relatively small amount of intentional impurities (dopants) that create electronic states inside the bandgap. Unwanted states are defects. Some of these are benign and some interfere with a desired behavior of the device. The vapor liquid solid (VLS) method 1 has attracted recent inter- est in growth of semiconductors for potential device use. VLS growth, because of the large surface of the wire, potentially suffers from a distribution of surface states as well as metal incorporation on the surface and in the lattice from the catalyst. All of these introduce defect states that cause trap-mediated nonradiative recombination, which reduces carrier lifetimes. Thus measurements of carrier lifetimes are a sensitive probe of material quality. Moreover for devices that rely on minority carrier transport, such as p n junction solar cells and bipolar junction transistors, the carrier lifetime is a critical parameter in device performance since it determines the diffusion length and the nonradiative losses in these devices. The carrier lifetime places limits on device design. Routine measurement of this parameter is important particularly in experimental materials. While carrier lifetimes have been mea- sured in VLS wires, 2 6 it is not routine. This is because the methods employed to date are complex. Most involve creating a single wire device. Using a device to measure lifetimes suffers from three problems: making a few single wire devices may not give statistically relevant results, the process of making any type of device can significantly change the material properties during processing, and it is time-consuming. For these reasons contactless lifetime measurement techniques were developed for thin films. In this work the classic photoconductance decay (PCD) method is extended to thin aggregated films of silicon VLS wires on a fused silica carrier. Al- and Au-catalyzed wires were grown using both annealed thin metal films and lithographically defined metal islands on Æ111æ Si wafers. Wires were grown to perform two experiments: one to assess the recombination contribution of the wire surface versus the bulk, and the other to assess the difference between Al- and Au-catalyzed wires (see Supporting Information for com- plete details). To assess surface effects, Au islands were defined lithographically before growth at 650 °C in SiH 4 , resulting in wires ranging in diameter from 392 to 730 nm (samples Au-Lith1,2,3). To compare catalyst types, wires were grown via annealed thin metal films of both Al and Au. The Al sample (Al-film) was grown at 490 °C and had a mean diameter of 144 nm, while the Au sample (Au-film) was grown at 600 °C and had a mean diameter of 125 nm. All Au samples had their catalyst removed in TFA Au etch and were then passivated with a high- temperature thermal oxide. The Al sample had its catalyst removed via HCl and then was passivated with a high-temperature thermal oxide. All samples were coprocessed whenever possible. The diameters reported above represent the silicon wire dia- meter after oxidation and are based on scanning electron microscope (SEM) measurements of oxidized wires combined Received: July 5, 2011 Revised: September 2, 2011 ABSTRACT: Carrier lifetimes of Si micro/nanowires grown by the vapor liquid solid method are measured using an extension of the classic contactless photoconductivity decay method. The samples measured consist of a thin aggregated film of oxide passivated wires on a fused silica carrier. Au catalyzed wires in the 392 730 nm diameter range are studied. Recombination in these wires is controlled by the surface or near surface effects, not bulk Au impurities. The lifetimes of Au- and Al-catalyzed wires of comparable diameter are measured. The Al wires are found to have slightly longer lifetimes than those grown with Au at a comparable diameter. Across all samples, the lifetimes measured range was from 0.2 to 1.0 ns. The surface controlled nature of the recombination measured implies larger diameter wires will offer better performance in devices that rely on minority carrier transport. KEYWORDS: Nanowire, lifetime, recombination, silicon, vapor liquid solid
Transcript
Published: September 22, 2011
r 2011 American Chemical Society 4282 dx.doi.org/10.1021/nl202279z |Nano Lett. 2011, 11, 4282–4287
LETTER
pubs.acs.org/NanoLett
Contactless Measurement of Surface Dominated Recombination inGold- and Aluminum-Catalyzed Silicon Vapor�Liquid�Solid WiresBrian A. Bryce,*,† Mark C. Reuter,§ Brent A. Wacaser,§ and Sandip Tiwari‡
†School of Applied and Engineering Physics and ‡School of Electrical and Computer Engineering, Cornell University, Ithaca,New York 14853, United States§IBM Research Division, T. J. Watson Research Center, Yorktown Heights, New York 10598, United States
bS Supporting Information
Semiconductor devices rely on material purity for theirproperties. The basis for nearly all semiconductor devices
is the introduction of a relatively small amount of intentionalimpurities (dopants) that create electronic states inside thebandgap. Unwanted states are defects. Some of these are benignand some interfere with a desired behavior of the device. Thevapor�liquid�solid (VLS) method1 has attracted recent inter-est in growth of semiconductors for potential device use. VLSgrowth, because of the large surface of the wire, potentiallysuffers from a distribution of surface states as well as metalincorporation on the surface and in the lattice from the catalyst.All of these introduce defect states that cause trap-mediatednonradiative recombination, which reduces carrier lifetimes.Thus measurements of carrier lifetimes are a sensitive probe ofmaterial quality. Moreover for devices that rely on minoritycarrier transport, such as p�n junction solar cells and bipolarjunction transistors, the carrier lifetime is a critical parameter indevice performance since it determines the diffusion length andthe nonradiative losses in these devices.
The carrier lifetime places limits on device design. Routinemeasurement of this parameter is important particularly inexperimental materials. While carrier lifetimes have been mea-sured in VLS wires,2�6 it is not routine. This is because themethods employed to date are complex. Most involve creating asingle wire device. Using a device to measure lifetimes suffersfrom three problems: making a few single wire devices maynot give statistically relevant results, the process of making anytype of device can significantly change the material propertiesduring processing, and it is time-consuming. For these reasons
contactless lifetime measurement techniques were developedfor thin films. In this work the classic photoconductance decay(PCD) method is extended to thin aggregated films of siliconVLS wires on a fused silica carrier.
Al- and Au-catalyzed wires were grown using both annealedthin metal films and lithographically defined metal islands onÆ111æ Si wafers. Wires were grown to perform two experiments:one to assess the recombination contribution of the wire surfaceversus the bulk, and the other to assess the difference betweenAl- and Au-catalyzed wires (see Supporting Information for com-plete details). To assess surface effects, Au islands were definedlithographically before growth at 650 �C in SiH4, resultingin wires ranging in diameter from 392 to 730 nm (samplesAu-Lith1,2,3). To compare catalyst types, wires were grownvia annealed thin metal films of both Al and Au. The Al sample(Al-film) was grown at 490 �C and had a mean diameter of144 nm, while the Au sample (Au-film) was grown at 600 �C andhad a mean diameter of 125 nm. All Au samples had their catalystremoved in TFA Au etch and were then passivated with a high-temperature thermal oxide. The Al sample had its catalystremoved via HCl and then was passivated with a high-temperaturethermal oxide. All samples were coprocessed whenever possible.The diameters reported above represent the silicon wire dia-meter after oxidation and are based on scanning electronmicroscope (SEM) measurements of oxidized wires combined
Received: July 5, 2011Revised: September 2, 2011
ABSTRACT: Carrier lifetimes of Si micro/nanowires grownby the vapor�liquid�solid method are measured using anextension of the classic contactless photoconductivity decaymethod. The samples measured consist of a thin aggregatedfilm of oxide passivated wires on a fused silica carrier. Aucatalyzed wires in the 392�730 nm diameter range arestudied. Recombination in these wires is controlled by thesurface or near surface effects, not bulk Au impurities. Thelifetimes of Au- and Al-catalyzed wires of comparable diameter are measured. The Al wires are found to have slightly longerlifetimes than those grown with Au at a comparable diameter. Across all samples, the lifetimes measured range was from 0.2 to 1.0 ns.The surface controlled nature of the recombination measured implies larger diameter wires will offer better performance in devicesthat rely on minority carrier transport.
4283 dx.doi.org/10.1021/nl202279z |Nano Lett. 2011, 11, 4282–4287
Nano Letters LETTER
with transmission electron microscope (TEM) measurements ofeach samples oxide thickness (see Supporting Information fordetails).
Following the surface passivation step, the wires wereremoved mechanically from their growth substrate into iso-propanol. The volume of the alcohol solution was reducedby centrifuging the solution and removing the upper fluidrepeatedly. The solution was then dispensed on a fused silicacarrier chip, and the alcohol was allowed to evaporate leavingthe wires as a thin aggregated film on the surface of the fusedsilica. This film resembles a pile of hay or needles microscopicallyand consists of hundreds of millions of wires (Figure 1). Theexperiment reported here measures the average lifetime of thisensemble.
In the past, PCD has been extended to work with III�Vmaterials7 which typically have significantly shorter lifetimesthan their group IV counterparts. Given the existing measure-ments in the literature,2�6 it was necessary to scale PCD stillfurther as lifetimes in the tens of ps are conceivable. Thisrepresents a significant technical challenge for the design of aPCD setup. These very short lifetimes in frequency domainrepresent wideband signals. This suggests that a PCD setupshould operate at the highest reasonable frequency so that thefractional signal/carrier bandwidth would be reduced. How-ever, this consideration neglects the availability of source poweras well as transmission line losses. Because of the particularequipment available to us, we chose to design for the X band;however a design for the K band would also be reasonable. Theexperimental design is shown in Figure 2 and is broadly similarto that of Kunst and Beck.8
In this experimental arrangement, the sample under study issuspended in a WR90 waveguide which is matched to the roomwith a microwave horn. This ideally allows the fused silica/wire
sample to be treated as a single impedance mismatch inside asemi-infinite waveguide. A 10 GHz microwave tone at a power of24 dBm is applied to the system through a �6 dB directionalcoupler used as a power splitter. The through port of thedirectional coupler is attached to a ferrite circulator which passesthe tone to a waveguide coupler. The continuous wave passesdown the waveguide and is partially reflected from the samplewith the rest of the microwave power passing out of the horn andinto the room. This arrangement makes the setup sensitive tomaterials outside of the waveguide as well. In particular, metallicobjects need to be kept far away from the end of the horn to avoidreflecting considerable microwave power back into the wave-guide. For this reason, dielectric mirrors were employed onfiberglass posts to route the pulsed laser onto the sample. Thispulsed laser injects excess carriers into the semiconductor wires,which changes their conductivity. This conductivity relaxes afterthe pulse. The relaxing conductivity continuously changes theimpedance mismatch in the waveguide. This produces anamplitude modulation (AM) on the reflected carrier wave. Thereflected wave is separated from the incoming wave by the ferritecirculator and passed to the radio frequency (RF) port of a phasequadrature (IQ) mixer. The reflected wave’s AM is detected bythe IQ mixer used in a homodyne arrangement and recordedusing an oscilloscope to give a measure of the voltage reflectanceas a function of time.
If the laser pulse is much shorter than the lifetimes, then thepulse width may be ignored. In the present experiment, aregeneratively amplified mode locked Ti�sapphire laser at800 nm with a pulse width of 100 fs is employed to achievethe carrier injection. Given the lifetimes measured, the 100 fspulse may be considered an impulse function. The peak pulsepower of the laser was 600 mJ, and the 1/e2 beam diameter ofthe beam was 25 mm after expansion. Thus the peak flux of an
Figure 1. SEMs of Au-Lith1: (a) as grown at 0�; (b) as grown at 90�;and (c) after the transfer process on the fused silica carrier. The wiresseen in (a) and (b) are growing in the Æ111æ direction. The normal of thewafer used was specified at 3.5� relative to the Æ111æ direction.
Figure 2. A 1 kHz repletion rate regeneratively amplified Ti�sapphirelaser at 800 nm generates 100 fs pulses with a peak power of 600 mJ perpulse. These pulses pass through a selectable ND filter and then areexpanded via a 5X Galilean beam expander. The pulse is routed viadielectric mirrors down a microwave horn and onto the sample, which issuspended in a WR90 waveguide. A microwave tone at 10 GHz at 24dBm is generated by a YIG-based CW sine source. This power is splitby a �6 dB directional coupler, and the coupled port is attached to theLO port of an IQ mixer and the through port to a ferrite circulator. Thecirculator forwards the power to the waveguide via a SMA to WR90coupler. The reflected power is AM modulated by the conductivitychange in the sample caused by the laser pulse. The reflected power issent to the RF port of the mixer via the circulator and demodulated intoits phase quadrature components: I and Q, which are recorded by anoscilloscope.
4284 dx.doi.org/10.1021/nl202279z |Nano Lett. 2011, 11, 4282–4287
Nano Letters LETTER
unattenuated laser pulse is nominally 10 GW/cm2. For silicon,at this flux and wavelength, linear absorption dominates.9,10
Typical PCD measurements attempt to relate the measuredreflectance signal to the absolute conductivity through a model.This approach makes sense in a sample that is well-definedgeometrically and is spatially homogeneous. The wire samplesare not spatially homogeneous on a micrometer scale. However,on the scale of the coupled microwave signal they are. Thisassures that the microwave sees a single effective impedancemismatch in the waveguide. However, this inhomogeneity doeschange the effective dielectric constant that the microwaveinteracts with. This complicates modeling. An effective mediumtheory11,12 could be used to construct a model of an aggregatedwire film if it was sufficiently characterized. However, such alengthy characterization would defeat the aim of a contactlessmeasurement of carrier lifetimes: simple rapid assessment ofmaterial quality. Because of this, a much more direct approach isemployed. The intensity of the laser light is varied to directlychange the relative carrier densities in the samples.
Consider a PCD experiment with a laser that creates Nelectron�hole pairs (EHP) at t = 0 in a semiconductor samplewith a reflectance transfer function Γ = g(σ), where σ is theeffective sample conductivity. If N. Na, Nd, then as the samplerelaxes toward equilibrium N = Ne = Nh, where Ne electrondensity, Nh is hole density, Na is the acceptor density, and Nd isdonor density in the sample. This condition is a high-levelinjection. In this case σ = eNμ, where μ is the sum of theelectron and hole mobilities. This means that in the highinjection limit, the conductivity is solely a function of N. Themobility also varies, but it too is only a function of N for a fixedsample at a fixed temperature.
This allows the definition of another reflectance transferfunction Γ = h(N). If this function were known, then a measureof reflectance would directly indicate the free carrier density inthe sample at any time as it relaxes. Thus a measurement ofreflectance would be a measure of the carrier lifetimes.
This transfer function can be measured experimentally. Con-sider two light pulses A and B with two different powers PA andPB, but otherwise identical incidences, on a sample. If theabsorption of the material with respect to light intensity is linear,then the ratio of themaximal number of EHP in thematerial fromthe two pulses is equal to the ratio of the light pulse powers, thatis: PB/PA = NB/NA at t = 0. This gives two points on the transferfunction. By using many different intensities, the entire functioncould be reconstructed except for a constant because of thedifferential nature of the mapping. This constant can eitherbe added to the function by an experimental estimate or couldbe removed from the experimental data numerically.
The method just described conceptually summarizes how toreconstruct the lifetimes using just reflectance data and lightpulses of different intensities. To be mathematically exact, such areconstruction needs to be differential in nature with a contin-uous differential variation in light pulse power. In practice this isof course not possible with a finite number of measurements. Asuitable method to achieve this same objective is to measure therelaxation time between pairs of pulses that inject different carrierdensities. This allows the average recombination rate to bemeasured between two relative carrier densities.
Again, consider the two pulses. To make things simpler,assume the recombination process is purely first order. In thatcase,N(t) =N0e
t/τ and then letNB =NAet1/τ from this. It follows
that τ = (t1/ln(NB/NA)) = (t1/ln(PB/PA)). In this thought
experiment, t1 is the time it takes the carrier density to relax toNB after pulse A. Equivalently, it is the time that it takes thereflectance Γ to relax to the amount of reflectance created bypulse B at t = 0. Thus, t1 can be measured by measuring thereflectance as a function of time for two light pulses of differentpowers. Meanwhile, the power of each pulse can be measureddirectly using a power meter. From these two measurements thelifetime τ of the sample can be extracted.
If the recombination process is not purely first order, then thelifetime measured is the time constant of the exponential neededto match the carrier density at t = 0 and t = t1. The differentiallysmall version of this method can build up the entire recombina-tion rate curve versus carrier density in the high injection levellimit. In terms of Hall�Shockley�Read (HSR) theory, this highinjection lifetime is: τ = τn + τp, where τn and τp are the electronand hole minority carrier lifetimes, respectively.13
In this procedure it was assumed that the materials’ absorptionis linear. This assumption should be checked experimentally foreach sample and any deviation accounted for.
Carrying out the procedure as described above results in afamily of reflectivity curves, as shown in Figure 3. To providelight pulses of different powers, a set of metallic neutral density(ND) filters in half order of magnitude increments was used. Tomeasure the average first-order lifetime, the time between pointsA and B for each pair was measured. This decay time representsthe convolution of the detector and sample responses. Thedetector response can be measured in either time or frequencydomain to obtain its contribution. The detector design used herehas a measured 3 dB bandwidth of better than 1 GHz. Thisinformation combined with a measurement of the pulse powerfor each ND filter completes the information needed to directlymeasure the relative decay time constant.
This method is sufficient to study the relative lifetime as afunction of injection level. To compare samples, it is necessary toknow the absolute injection level of the sample. This can also bedirectly measured by combining measurements of the sample’sactive volume with measurements of the samples absorption.
Figure 3. A typical family of reflectance traces from the experimentalapparatus, in this case for sample Au-Lith2. The pulse attenuation causedby each ND filter is known via a power meter measurement. The ripplesare from the mixed up frequency component which is not completelyfiltered. To measure the time between A and B, each curve is fit to anexponential decay, and the time offset calculated.
4285 dx.doi.org/10.1021/nl202279z |Nano Lett. 2011, 11, 4282–4287
Nano Letters LETTER
A simple estimate of the active volume may be made byknowing the wire density and the sample size before the transferprocess and by assuming some transfer efficiency. A transferefficiency of 100% places an upper bound on the active volume.Direct inspection of the two-dimensional coverage of the carrierfused silica, combined with SEM and TEM measurements of thewire diameter, provides a lower bound estimate of the activevolume (any covered area has at least one wire layer). A moreaccurate estimate of the volume can be made by combiningSEM/TEM observations of the wires and the wire film withprofilometry to estimate the film porosity, wire size, and filmthickness. If higher accuracy is needed, thenmeasurements of thesample volume might also be made using a crystal mass sensor.For the present work, a combination of SEM, TEM, andprofilometry measurements, along with the bounding estimatesdescribed, were used to calculate the active volume.
In the present experiment, a Shimadzu UV-3101PC spectro-photometer was employed to measure the reflected and trans-mitted power and thus the absorption. The transmitted powerwas compared with the measured transmitted power of the high-power experimental laser with a Coherent FieldMax II/PM3thermopile sensor to verify the linear absorption assumption foreach sample, with any deviation added to the error in thereported lifetime.
These procedures were carried out on five samples. The resultsare summarized in Table 1. Meanwhile the samples lifetimes arecompared at a nominal injection level of 1� 1018 EHP/cm3 inFigure 4. The samples are compared at this relatively high injectionlevel because of the strong p-type doping of Al-catalyzed wires andthe high injection level requirement.
The errors reported and plotted in the diameter representboth the spread in the measured diameter across the ensemble aswell as the taper of the wires from the base to the tip. The taper isa result of conventional CVD processes occurring in parallel withthe VLS type growth. This conventional CVD growth occurs onthe sidewalls of the wires as well as the spaces between them witha rate and quality similar to the conventional planar CVD thatwould occur under identical reactor conditions.
The errors reported and plotted for the lifetime represent theerrors from the time base and trigger jitter of the oscilloscope,absorption measurements, and power detector measurements aswell as the random error of the detected signal.
From the data, several observations are immediately clear.First, in the three samples meant to compare the effect ofdiameter on lifetime, a strong dependence on the diameter isobserved. Second, Al-catalyzed wires are not significantly super-ior to their Au counterparts. Third, although the injection levels
used are high, Auger processes may be ignored. (The Augercoefficient for Si is nominally 1 � 10�30 cm6/s, resulting in aneffective lifetime of 10 ns at an injection level of 1 � 1019 EHP/cm3.)13 Thus, we may consider only defect-mediated recombina-tion processes both at the surface and in the bulk.
Consider the three coprocessed Au samples with lithographi-cally defined islands (Au-Lith1�3). In the data shown in Figure 4,we see a very strong dependence on diameter. This stronglyindicates that surface or near surface processes are the dominantsources of recombination even in these relatively large VLS wires.In fact these data completely preclude the commonly heldview that bulk Au contamination is the principle reason for therelatively low lifetimes of VLS materials. Because these sampleswere grown together and processed together in all steps, if bulkAu was the dominate recombination mechanism, then all of theirlifetimes would be approximately equal, which is not the case.
The cause of the surface area-dependent recombination in thesesamples is not clear from these data however. There are severalpossible and indistinguishable sources of recombination in thesesamples that may appear as surface, like terms, traditional surfacerecombination at the Si/SiO2 interface, recombination fromenvelopedAunear the surface, and lower lifetime sidewallmaterial.
Traditional surface recombination caused by defects at theSi/SiO2 interface is certainly present and unavoidable in these
Figure 4. Comparison of carrier lifetimes across samples at a nominalinjection level of 1 � 1018 EHP/cm3. The lifetimes seen are linearlyinterpolated for this injection level using the data in Table 1, namely therelative ND-based carrier lifetime and the nominal EHP injection level.
4286 dx.doi.org/10.1021/nl202279z |Nano Lett. 2011, 11, 4282–4287
Nano Letters LETTER
samples. This mechanism should be approximately equal per areaon each sample given they were all passivated with the same high-temperature thermal oxide. If this were the only mechanism, thedata should fit a simple two parameter theory with a surfacerecombination velocity and a bulk lifetime. For the data inFigure 4 such a theory is a poor fit. Attributing all of therecombination in each wire sample to surface processes can givean upper bound on the surface recombination velocity. Doingthis for the data in Figure 4 results in surface recombinationvelocities ranging from 2� 104 to 4� 104 cm/s. These values aresignificantly in excess of what is expected for a thermal oxidepassivation (∼102 cm/s).7 Both of these facts are suggestive thatother mechanisms are likely present.
Au is known to migrate along wire sidewalls in UHV con-ditions.14 The exact Au sidewall coverage appears to be a functionor reactor conditions.15 During a typical VLS wire growth,regular CVD processes are occurring in parallel with the VLSmechanism resulting in wire taper and the accumulation ofregular CVD material on the wire sidewalls. While this sidewallAu may remain on the surface at all times (as the catalyst does), itcannot be ruled out that the sidewall growth envelops some ofthis Au with Si. As the wire diameter decreases, this could create aAu concentration greatly in excess of the solid solubility at thetemperature used. More importantly it would appear as a surface-like term with the Au localized near the surface. This Au wouldnot be removed by the TFA Au etch used to remove the catalystbecause of the Si covering it.
The surface term could also arise simply from the radialgrowth. The material that grows on the sidewall, althoughclearly epitaxial given the faceted nature of the resulting wires,may simply be more defective in terms of crystalline defectsthan the VLS core of the wire.16 Because the sidewall materialgrows at a rate that is essentially a constant radial thickness pertime, there is a greater ratio of sidewall to bulk material on wiresthat have a smaller diameter. Thus any deleterious effect of thismaterial would affect wires of smaller diameters more. We arenot able to distinguish the sidewall material from the core wirein TEM observations; however, we are able to place a bound onthe thickness of this material based on the wire taper (Table 1).This material thickness has no thickness at the wire tip and mayincrease to a thickness as great as the reported upper bound atthe wire base.
Now consider the two film based samples (Au-film and Al-film). From the data, we see there is an improvement for Al-catalyzed wires over their Au grown counterparts. However, onemust consider the added complexity of using in situ Al evapora-tion when deciding between Al- and Au-catalyst types. The datahere may understate the advantage of Al-catalyzed wires. Asshown in the diameter dependence experiment, surface effectsare dominant, not bulk effects, so if these surface effects could befully mitigated, then Al-catalyzed wires could indeed be superior.However, it is currently difficult to grow large diameter Al wires.This means that suppressing these surface effects is a significantchallenge for the Al�Si VLS system. Until this is accomplished,simply growing larger diameter Au wires is likely to achieve betteroverall lifetimes.
It is tempting to compare the Au-film and Au-Lith samples,however they are not directly comparable. The Au-film samplewas grown at lower temperatures than the Au-Lith set. Thisresults in a lower growth rate of both the wire and the sidewallmaterial. In the Au-film sample more of the sidewall material wasoxidized in the passivation step, and the passivation oxide was
grown at higher temperatures. These differences prevent a closecomparison between the Au-film and Au-Lith samples. Howeverthe lifetime of the Au-film sample is longer than onemight expectbased on the Au-Lith data. We note that over half of the wirelength in the Au-film sample is free of sidewall material, whichwhen combined with the Au-Lith data is suggestive that thismaterial may be playing a prominent role in the recombination inthis system.
Lifetime determination is important for assessing electronicquality of materials for applications where the bipolar charac-teristics are important, such as in minority carrier devices,including p�n junction photovoltaics. A simple extensionof the PCD method allows the determination of carrier lifetimesof VLS wires. The procedure for measurement is simple: transferthe material to a transparent insulator; measure the microwavereflectance transient to at least two light pulses of differentintensities and the power of the pulses; obtain the average first-order lifetime for the mean injection level; obtain the absolutecarrier density by measuring the samples absorption and activevolume.
We find that even with the use of reasonable passivationtechniques that surface and near surface effects in Au-catalyzedwires are the dominant reason for the very short lifetimes in VLSmaterials. We also find that Al-catalyzed wires may have a slightadvantage over their Au counterparts; the effect however is notdramatic. We note the slight advantage of Al over Au is of limitedvalue, if a larger diameter Au catalyzed can be used in place withequal efficacy.
’ASSOCIATED CONTENT
bS Supporting Information. TEM observations, completesample growth and preparation procedures. This material isavailable free of charge via the Internet at http://pubs.acs.org.
We thank Andrew Davis and Chris Schaffer for the use of theirlaser. We thank Jared Hertzberg and Tchefor Ndukum for usefuldiscussions on microwave system design. This work was sup-ported by the Cornell Center for a Sustainable Future.
’REFERENCES
(1) Wagner, R. S.; Ellis, W. C. Trans. Metall. Soc. AIME 1965, 233(6), 1053.
(2) Putnam, M. C.; Turner-Evans, D. B.; Kelzenberg, M. D.;Boettcher, S. W.; Lewis, N. S.; Atwater, H. A. Appl. Phys. Lett. 2009,95 (16), 163116.
(3) Kelzenberg, M. D.; Turner-Evans, D. B.; Kayes, B. M.; Filler,M. A.; Putnam,M. C.; Lewis, N. S.; Atwater, H. A.Nano Lett. 2008, 8 (2),710.
(4) Strait, J. H.; George, P. A.; Levendorf, M.; Blood-Forsythe, M.;Rana, F.; Park, J. Nano Lett. 2009, 9 (8), 2967.
(5) Allen, J. E.; Hemesath, E. R.; Perea, D. E.; Lensch-Falk, J. L.; Li,Z. Y.; Yin, F.; Gass, M. H.; Wang, P.; Bleloch, A. L.; Palmer, R. E.;Lauhon, L. J. Nat. Nanotechnol. 2008, 3 (3), 168.
4287 dx.doi.org/10.1021/nl202279z |Nano Lett. 2011, 11, 4282–4287
Nano Letters LETTER
(7) Yablonovitch, E.; Gmitter, T. J. Solid-State Electron. 1992, 35 (3),261.(8) Kunst, M.; Beck, G. J. Appl. Phys. 1986, 60 (10), 3558.(9) Bristow, A. D.; Rotenberg, N.; van Driel, H. M. Appl. Phys. Lett.
2007, 90 (19), 191104.(10) Green, M. A.; Keevers, M. Prog. Photovoltaics 1995, 3 (3), 189.(11) Granqvist, C. G.; Hunderi, O. Phys. Rev. B 1978, 18 (4), 1554.(12) Tinga, W. R.; Voss, W. A. G.; Blossey, D. F. J. Appl. Phys. 1973,
44 (9), 3897.(13) Schroder, D. K. IEEE Trans. Electron Devices 1997, 44 (1), 160.(14) Hannon, J. B.; Kodambaka, S.; Ross, F. M.; Tromp, R. M.
Nature 2006, 440 (7080), 69.(15) Dailey, E; Madras, P.; Drucker, J. J. Appl. Phys. 2010, 108 (6),
064320.(16) Wacaser, B. A.; Reuter, M. C.; Khayyat, M.M.;Wen, C.; Haight,
R.; Guha, S.; Ross, F. M. Nano Lett. 2009, 9 (9), 3296.
